Algorithms Will Make You Sad

algorithmsUsing algorithms to predict preferences and make recommendations to individual consumers is the hottest thing in marketing. Netflix recommends movies, Pandora recommends music, and wine websites increasingly rely on “if you like that, you’ll like this” to lure customers into making a purchase. And it’s understandable why. Instead of enduring the trouble of actually communicating with someone and dealing with the idiosyncrasies of their subjectivity, you can let the machines do the work of sorting through data. All the human work is done up front. Once the algorithms are written, sellers can set back and reap the rewards of matching buyers with the products they like. And buyers don’t have to worry about whether the salesperson understands them or not. The machine has all their past purchases to rely on. It’s all very efficient, impersonal, and reduces the complexity of transactions.

The problem is that human subjectivity doesn’t work like that. Our preferences aren’t consistent and our likes and dislikes are not based on general properties of things but instead on what makes them unique. I love black coffee and bitter greens, but that doesn’t mean I like a bitter wine. I adore a good Baume de Venise but that doesn’t mean I want sugar in my Chardonnay.

Commenting on her experience taking a “palate quiz” provided by Tastry, a start-up that analyzes wine at the molecular level to make their recommendations, wine writer Esther Mobley laments:

I downloaded the app and took the survey. It presented a series of flavors and aromas to me and asked whether I was pro, neutral or anti. I answered honestly (I’m pro-black olive, anti-sour candy, neutral on the smell of cigar tobacco) and waited eagerly as the app calculated my results.

The first wine listed, clocking in at a 99% match to my idiosyncratic palate, was Barefoot Sweet Red ($5.82). My heart sank. That is not a wine that I would like to buy.

Mobley thinks we buy the label, the story, the values of the product—what we think is cool—rather then just flavor. No doubt that is true. But even if we focus on flavor alone,  I doubt that our preferences can be measured or precisely determined. Aesthetic experience is valuable because it is idiosyncratic and radically particular. It’s that part of life where inconsistency is welcomed and expected. Algorithms threaten to rob us of that dimension of experience.

I really don’t know if recommendation engines work or not. I don’t have data to support an argument on way or the other. But I know they don’t work for me. I simply ignore them.

It may well be that they work for casual consumers who enjoy the ease and simplicity of being habituated. But for someone serious about film, music, or wine appreciation, they will make you sad.


  1. People in the wine business almost always make the mistake of believing that consumers buy “what they like.” No doubt that is true—people don’t buy what they don’t like, after all. But many wine consumers are explorers, and look for new ways to wine. For every Rombauer Chardonnay devotee, there is a consumer who loved the Catena Malbec they had in the restaurant last night, and wants to know what else they could find that would be like that. Note that they are not looking to buy a case of the Catena Malbec. But they have “discovered” another area of wine to explore, and want to know who else makes Malbec, what else does Catena make, and does Chile make wines that might be as interesting as those in Argentina?

    This is why brand loyalty is such a rare commodity in wine. And Esther is correct–if you give people a good reason to investigate a new kind of wine, they will be predisposed to taste it with an open mind and a receptive palate.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.